Depends on when they had their MCU repaired/replaced. If they got one with the 64GB eMMC then they should be fine. If they got another one with a 8GB eMMC they would still be at risk.
Correct, recalls apply to salvage vehicles as well.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
NHTSA asks Tesla to recall 158,000 [now 135,000] vehicles for eMMC failure. Voluntary Recall issued
- Thread starter Cheburashka
- Start date
Depends on when they had their MCU repaired/replaced. If they got one with the 64GB eMMC then they should be fine. If they got another one with a 8GB eMMC they would still be at risk.
Correct, recalls apply to salvage vehicles as well.
Actually you can't. They said they would announce details of the reimbursement program in February. But given the NHTSA request for a recall I could imagine they might delay that now.
ucmndd
Well-Known Member
Nice!
I have my hat at the ready, although I suppose there’s some weasel room RE what constitutes “mandatory” right now..
This hat is nasty, gonna need lots of hot sauce...
68882
Member
Spinning up an assembly line in China to build 200k MCU1 replacements is not the problem. It’s the 2? million hours of admin & shop time needed to swap the units. So what if you setup a program to train 5,000 mechanics to do this, later skim those folks for full time employment .
In the letter to Tesla, the NHTSA estimates that the MCU failure takes an average of only 5-6 years. I bought a used Model S from Tesla (so still under warranty) and my MCU failed after 4.6 years and 36k total miles. It was replaced under warranty (with a refurbed MCU1) but since the odds of it failing again are 100% according to NHTSA, I'm curious what they are going to do permanently fix this. Tesla has a history of taking the cheap and easy way out and now they are going to have to pay for this behavior.
ucmndd
Well-Known Member
For many months now, they have been replacing the Tegra daughter boards in MCU1 with a new part that has a 64GB (vs 8GB) chip. This is the permanent fix.
ucmndd
Well-Known Member
Almost 13 hours per car? That’s insanely high.
1.3 hours per car is much closer to reality.
What about cars like mine where I've been having MCU issues, but they told me my eMMC is fine? Its the CPU that is overloaded. I'll take a brand new MCU2 for free!
My car's had various documented "hey my MCU got super wonky, here's a picture of it not working" "hey the update is wedged" "hey my screen is black and my AC is stuck on and my turn signals don't work"; each time service has "sorted" the problem remotely or patted me on the hand and said they understand my feelings.
But I've still got the same dementia infested MCU that takes 30-300 seconds to realize it is a car, not a tablet or a drone controller or a video game system when I get in it to drive it. Doesn't matter if it's been off overnight or if I'm just getting out of the car for 30 seconds to help a little old lady cross the street, every time I get back in I get the blank screen of anticipation or the ghostly white T of quality on the center screen for a random amount of time.
Thus far it always realizes it is a car, and does car things. I've taken the car back several times and they reassure me that everything's fine and they will replace the MCU if their script indicates that they should.
So, the warranty isn't actually getting the thing replaced, unless it is so totally dead that they can't just revive it with a force-flash in the shop.
I suspect a recall will actually replace the physical unit in the car with one that works.
68882
Member
The actual replacement work is low, acquiring staff, training, sending letters, negotiations with customers, tracking down cars, making the parts, moving inventory around. There is a whole hidden effort that goes into the process before the car door is opened by the guy holding the parts.
I thought their current software already reduced the logging as much as they can. It doesn't fix the problem. It just slows down how quickly it fails.
Of course it won't help me given I'm on firmware from 2018. If my MCU fails, I'll be forced to upgrade to current software which WILL reduce my supercharging speed by nearly half and likely will software cap my maximum battery capacity and power.
If Tesla doesn't offer any discounted path to MCU2, I'm going to have them do the recall on my car, and then when there picking the car up, set up an appointment for MCU2.
studioloft
Gavin
This is exactly the problem. Look at the global semiconductor industry right now. AMD came out with new processors recently. NVidia came out with new graphics cards. Both companies have released the products in the thousands where /userconsumer demand is hundreds of thousands. They simply can't manufacture semiconductors fast enough with the given state of the world.
I’m just over 4 years and 56k miles. So far so good on my MCU1 but I will be upgrading to MCU2 this spring regardless. I have FSD. Might as well upgrade the car to take full advantage. Beats buying new.
Understand, I am talking about the financial implications of this for Tesla. They already said they were paying for the fix and reimbursing people, like us, that paid for a replacement.
And these older products may no longer be produced, or they're making something that's somewhat like the original product but has different timing or voltage characteristics because they took some design and moved it to another fab. At the very low or high end of the volume production this isn't a big deal, but 200,000 units is actually below the threshold for nvidia to really work super hard to accommodate you. And the problem isn't just (or even *mostly* the CPU; if there are any other custom ASICs on the board you'll need those fabbed as well, and ideally the new ones will be identical to the old ones in every way, which is not trivial)
This isn't like making bolts or airbags; IC's need extremely expensive fabs to be made and those fabrication plants are constantly in flux. The fabs make ICs of a particular sort (voltages and timings are dictated by the feature size of the fab) and there just aren't that many fabs out there with spare capacity.
If I had the task of rolling through 1000 repairs a month, I'd gamble on being able to do it with refurbishing existing units. If tesla was actually throwing these out in the early phases of this issue, it indicates they didn't understand their liability about this whole process. Those cores are pretty valuable.
If I had the task of rolling through 30,000 repairs a month and had a hard deadline to finish, I'd tell mgmt they're better off making the existing MCU2 talk to an analog tuner / remote instrument cluster and put a feature flag that disables "mcu2" features.
I'd also have hard supply contracts lined up with whoever can make all the BOM for the MCU1 as the alternative.
It is solvable, but it is not good.
Last edited:
ucmndd
Well-Known Member
I understand and agree with your assessment, but still think 13 hours per car is astronomically high. They are not reinventing a service and parts distribution infrastructure from scratch for this (although we can certainly discuss whether or not they should
). They have a million cars on the road at this point and growing by half a million a year - ~150,000 service appointments spread out over a couple years is a considerable extra volume, but not a "holy sugar this changes everything" volume.
